Search code examples
statisticsregression-testinggretl

Gretl - test alpha and beta

I have done an ordinary least squared with an excel data set. Now I want to test α against the value of 0 and β against the value of 1 using an error probability of 0.05%.

How to do this in gretl?

I appreciate your answer!!!

Solution

  • I edit your post to add a gretl flag, hope we'll have more gretl users in SO too.

    Now to answer to your question, you need to deal with general linear restriction in your model and computing the related F-test.

    So to do it with gretl, you can do something like this using gretl scripting language hansl :

    open murder_rates
ols executions const income lfp southern --quiet

restrict
b[income]=0
b[lfp]=1
end restrict

    And we have the following result

    ## Restriction set
##  1: b[income] = 0
##  2: b[lfp] = 1

## Test statistic: F(2, 41) = 29633.7, with p-value = 1.6337e-65

## Restricted estimates:

##              coefficient   std. error   t-ratio    p-value 
##   ---------------------------------------------------------
##   const       -53.0056      0.370949    -142.9     3.29e-59 ***
##   income        0.00000     0.00000       NA      NA       
##   lfp           1.00000     0.00000       NA      NA       

##   Standard error of the regression = 2.4606

    If you want to do it with easily with R check the car package (linearHypothesis function)